Hepcidin levels predict nonresponsiveness to oral iron therapy in patients with iron deficiency anemia

Authors


  • Conflict of interest: Consultant, Luitpold Pharma

  • Author Contribution DBB, DM, and TAK designed the research and analyzed the data; DM also performed statistical analysis; LTG designed the research, participated in the analysis of the data, and the writing of the manuscript; AH analyzed the data and generated the initial draft of the manuscript. DBB and TAK are employees of Luitpold Pharmaceuticals, the Sponsor of the study. AH is an employee of St. John's University. DM and LTG are consultants for Luitpold Pharmaceuticals.

Correspondence to: Lawrence Tim Goodnough, Stanford University, 300 Pasteur Drive Room H-1402, M/C 5626, Stanford, CA94305. E-mail: ltgoodno@stanford.edu

Abstract

Levels of hepcidin, a major regulator of iron homeostasis, may identify patients with iron deficiency anemia (IDA) who will not respond to oral iron therapy. In this study, IDA patients underwent a 14-day trial (run-in) course of ferrous sulfate therapy. Nonresponders (Hgb increase <1 g/dL with 67% compliance rate) were randomized to IV ferric carboxymaltose (FCM; two injections of 750 mg) or further oral iron for 14 days. Screening hepcidin levels were 38.4 versus 11.3 ng/mL, P = 0.0002 in nonresponders versus responders to a trial of oral iron. Hepcidin of > 20 ng/mL, showed sensitivity of 41.3%, specificity of 84.4%, and positive predictive value of 81.6% for predicting nonresponsiveness to oral iron. PPVs for ferritin> 30 ng/mL or transferrin saturation (TSAT)>15% were 59.2 and 55%, respectively. Negative predictive values for hepcidin, ferritin, and TSAT were 46.3, 22.7, and 19.7, respectively. FCM versus oral iron showed Hgb increases of ≥1 gm/dL in 65.3% versus 20.8% (P < 0.0001) and Hgb increases of 1.7 ± 1.3 versus 0.6 ± 0.9 g/dL (P = 0.0025), respectively. We conclude that hepcidin predicts nonresponsiveness to oral iron in patients with IDA and is superior to TSAT or ferritin for this purpose. Nonresponse to oral iron therapy does not rule out IDA, since two-thirds of patients subsequently responded to intravenous iron. Am. J. Hematol. 88:97–101, 2013. © 2012 Wiley Periodicals, Inc.

Introduction

Worldwide, iron deficiency anemia (IDA) is the most common nutritional deficiency [1] The etiology of IDA may result from inadequate intake of iron, impaired absorption, or losses of iron from various conditions [e.g., menstrual or gastrointestinal (GI) blood loss]. Treatment options for IDA include supplementation with oral iron or intravenous (IV) iron therapy. Although oral iron has been the initial choice to treat patients with IDA, there are patients who do not tolerate the side effects of oral iron or who do not experience an optimal response for management of IDA within a prescribed time period. In many patients, inflammation or other co-existing conditions inhibit the ability of iron to be absorbed from the GI tract or released effectively from storage iron to hematopoietic precursors [2]. Parenteral administration of iron has been used in these patients when oral iron is not a feasible or effective option.

The homeostasis of iron in the body is regulated in part by a peptide hormone called hepcidin, which is secreted by hepatocytes [2]. Hepcidin acts on the transmembrane iron efflux transporter ferroportin [3]. This transporter is present on the basolateral surface of enterocytes, and is also present in macrophages, hepatocytes, and placental cells [4, 5]. The cell surface expression of ferroportin permits the movement of intracellular iron into plasma [6]. When hepcidin is bound to ferroportin, this transporter is internalized and degraded [4]. Consequently, ferroportin can down regulate intestinal iron absorption as well as the release of iron from macrophages and iron stores within hepatocytes [4].

The laboratory definition of absolute iron deficiency has traditionally been based on low serum iron, low percent transferrin saturation (TSAT), and low ferritin [7]. However, laboratory results can be less sensitive and/or specific in patients who have co-morbid disease with inflammation, particularly when iron deficiency co-exists with inflammatory conditions that lead to iron sequestration and hypoferremia [8]. Entrapment of iron occurs within macrophages and within GI epithelial cells, impairing GI absorption. This redistribution of iron during inflammation is due to cytokine-stimulated over production of hepcidin [9]. The identification of iron-restricted erythropoiesis due to absolute iron deficiency and/or iron sequestration is important for management of anemia [10].

As hepcidin is the major regulator of iron, there may be a benefit in utilizing this hormone as a biomarker for clinical assessment of patients with IDA. We conducted a sub-study from a previously reported phase III clinical trial [11] in which screening levels of hepcidin were collected to identify patients with IDA who may respond adequately to a course of oral iron or not. In addition, the randomization to two forms of iron therapy (i.e., ferric carboxymaltose (FCM) vs. oral iron) was analyzed for changes in hepcidin levels from baseline (Day 1) to Day 35.

Methods

Hepcidin measurements were obtained from patients enrolled in a previously reported randomized, multicenter, controlled trial comparing the efficacy and safety of FCM in patients with IDA [11]. Eligibility for this trial included patients ≥ 18 years of age with hemoglobin ≤11 g/dL and IDA of any etiology with ferritin ≤100 ng/mL, or ≤300 ng/mL when TSAT was ≤30%. The study design for this trial is illustrated in Supporting Information Fig. S1. Patients who met the inclusion criteria received a 14-day course of oral iron (“oral iron run-in”-ferrous sulfate 325 mg three times daily for 14 days). The 14 day run in criteria were selected based on a review of trials of FCM in patients with IDA in the postpartum period or IDA and heavy uterine bleeding (HUB) or inflammatory bowel disease [12-15]. It was observed that patients who achieved a <1 g/dL increase in 14 days were less likely to achieve satisfactory responses to oral iron at the end of the respective studies than those who achieved a ≥1 g/dL increase in 14 days. Patients who had an adequate response to the oral iron run-in (operationally defined as a hemoglobin increase of at least 1 g/dL in 14 days) were categorized as “responders” and were therefore not included in the aforementioned randomized trial comparing the efficacy and safety of FCM in patients with IDA [11]. The basis of the current substudy evaluated the utility of hepcidin for predicting responsiveness to oral iron, based on a comparison of the responders to oral iron (who were not included in the randomized trial [11] to the oral iron nonresponders (who were included in the randomized trial).

Those that did not have at least a 1 g/dL increase to the oral iron run-in in 14 days (in which patients had a 67% compliance based on pill count) were classified as “nonresponders” and randomized to treatment with either FCM from (Group A) or continuation of oral iron (Group B) for another 14 days. FCM (Group A) was administered at a dose of 15 mg/kg to a maximum 750 mg per dose administered intravenously on Day 0 and 7. Oral iron was given as ferrous sulfate 325 mg three times a day for an additional 14 days starting on Day 0. Hemoglobin levels and markers of iron status were assessed up to 35 days and measures of safety and tolerability (including a composite cardiovascular safety endpoint independently adjudicated) were assessed up to 120 days.

Selection of screening (Day 15) hepcidin levels to predict if patients would respond adequately to oral iron was based on an initial group (Analysis Group I) of 44 patients (22 responders and 22 nonresponders). The assessment of hepcidin levels occurred in a subset of study sites based on site willingness to collect an additional blood sample for this purpose. Selection of subject samples for assessment was performed by one author (TK) who sequentially selected 44 subjects based on response to the oral iron run-in without any knowledge of any other laboratory or demographic characteristics (i.e., by simply selecting patient identifier numbers from a list that contained no other subject information). The initial sample size for screening hepcidin levels in Analysis Group I was based on an estimated difference of mean 60 ± 60 ng/mL between responders and nonresponders [16]. Data from Analysis Group I was analyzed in a scatter plot (≥1 g/dL change in hemoglobin [yes/no] vs. baseline hepcidin value) to identify a hepcidin value for further analysis. A total of 240 patients (Analysis Group II), selected as described above, was evaluated subsequently to determine if hepcidin was a reliable predictor of nonresponse to oral iron.

Hepcidin levels were also assessed after the oral iron run-in (Day -1; responders and nonresponders) and after treatment with IV iron or oral iron post-randomization (Day 35; nonresponders only since responders were not randomized). Serum hepcidin levels were measured using a commercially available competitive enzyme-linked immunosorbent assay (C-ELISA). Samples were collected by the investigator at the site, stored by a contract research organization (Covance, Indianapolis, IN) and sent to a commercial laboratory (Intrinsic LifeSciences, La Jolla, CA) for analysis.

The data from all patients in this substudy for which hepcidin measurements were available were included for analysis. No imputation of missing data was performed. All statistical comparisons were made with a 0.05 Type I error for two-sided tests. Tests for proportions were performed with Fisher's exact test or its generalized version for>2 categories. Mean differences were assessed with the t-test (i.e., one-way analysis of variance for two groups), assuming equal variances. Correlations were estimated with Pearson's product-moment method. Logistic regression was used to evaluate the relationship between response to oral iron and screening values of hepcidin, ferritin, and TSAT. The following definitions were used to calculate sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Sensitivity =100* (# nonresponders correctly predicted)/(# observed non-responders).

Specificity = 100* (# responders correctly predicted)/(# observed responders).

PPV = 100* (# non-responders correctly predicted)/(# predicted non-responders).

NPV = 100* (# responders correctly predicted)/(# predicted responders)

Results

Forty-four patients (Analysis Group I, 22 nonresponders and 22 responders) were studied in the initial analysis. Screening values for hepcidin, hemoglobin, ferritin, iron, total iron binding capacity, and TSAT were assessed. A comparison of nonresponders to responders for screening hepcidin and iron indices are detailed in Table 1. For this initial analysis group, screening hemoglobin levels were higher in nonresponders versus responders (P = 0.0012). Screening hepcidin levels were significantly higher in nonresponders versus responders (33.2 vs. 8.7 ng/mL, P < 0.004). Mean ferritin levels were also significantly higher in the nonresponders versus responders (26.4 vs. 7.2 ng/mL, P = 0.03). Differences in screening TSAT (%) levels did not reach statistical significance.

Table 1. Screening Hepcidin and Iron Indices of Nonresponders and Responders to 14 Day Oral Iron Run-In (Analysis Group I)
Screening analyteDescriptive statisticNonresponder (Hgb ▵ <1 g/dL) (N = 22)Responder (Hgb ▵ ≥1 g/dL) (N = 22)P-value
Hepcidin (ng/mL)Mean (SD)33.2 (36.6)8.7 (7.1)0.0036
 Median20.87.8 
 Min, Max0.0, 139.30.0, 20.2 
Hemoglobin    
(g/dL)Mean (SD)10.1 (0.8)9.1 (1.1)0.0012
 Median10.19.0 
 Min, Max7.6, 11.67.3, 11.1 
Ferritin    
(ng/mL)Mean (SD)26.4 (39.7)7.2 (9.4)0.0326
 Median10.73.9 
 Min, Max2.2, 176.71.0, 38.2 
Iron    
(microg/dl)Mean (SD)45.3 (35.3)30.8 (47.3)0.2567
 Median28.015.5 
 Min, Max8, 1448, 234 
TIBC    
(microg/dl)Mean (SD)326.5 (64.4)361.3 (45.8)0.0453
 Median316.5375.0 
 Min, Max200, 461269, 444 
TSAT (%)    
 Mean (SD)14.3 (10.6)8.5 (12.0)0.0950
 Median8.54.0 
 Min, Max2, 422, 59 

A scatter plot of screening hemoglobin versus screening hepcidin with response to oral iron for Analysis Group I is depicted in Fig. 1. This scatter plot indicated that only 1 of 22 subjects with screening hepcidin values >20 ng/mL were responders, and this cutpoint was selected for further evaluation in the larger Analysis Group II.

Figure 1.

Scatter plot of screening hemoglobin (y-axis) versus hepcidin (x-axis) with response to oral iron identified. This scatter plot shows the hemoglobin level and respective hepcidin level (ng/mL) for the initial 44 subjects assessed in the substudy (Analysis Group I) after the 14 day oral iron run-in. Patients in solid circles were defined as responders and patients in open circles were defined as nonresponders. Response was defined as a hemoglobin increase of >1 g/dL after run-in.

Based on Analysis Group I, the predictive value of hepcidin level>20 ng/mL to predict nonresponsiveness (increase in hemoglobin of ≤1 g/dL) to oral iron was tested in a larger analysis group (i.e., Analysis Group II) of 240 subjects (150 nonresponders and 90 responders, also including the 44 subjects in the initial Analysis Group I). Demographic and baseline characteristics of Analysis Group II are summarized in Supporting Information Table SI. Two hundred twenty six (>90%) of the patients were female. One hundred and eight (43%) and 35 (14%) had HUB and GI bleeding, respectively. A comparison for screening hemoglobin, hepcidin and iron indices of nonresponders to responders in Analysis Group II is provided in Table 2. Screening hemoglobin was higher in nonresponders versus responders, (10.1 vs. 9.3 g/dL, P < 0.0001). Patients within Analysis Group II reflected greater hemoglobin changes in responders versus nonresponders after the 14 day oral iron run-in (1.8 ± 0.6 vs. 0.2 ± 0.4 g/dL, P < 0.0001, respectively). Screening hepcidin levels were significantly higher in nonresponders versus responders (38.4 vs. 11.3 ng/mL, P = 0.0002). Screening ferritin values were also higher in nonresponders versus responders (31.9 vs. 12.2 ng/mL, P < 0.03).

Table 2. Screening Hepcidin and Iron Indices of Nonresponders and Responders to 14 Day Oral Iron Run-In (Analysis Group II)
AnalyteDescriptive statisticNonresponder (Hgb ▵ <1 g/dL) (N = 150)Responder (Hgb ▵ ≥1 g/dL) (N = 90)P-valuea
  1. aFrom 1-way ANOVA.
Hepcidin    
(ng/mL)Mean (SD)38.4 (66.7)11.3 (19.0)0.0002
 Median14.706.40 
 Min, Max0.0, 540.70.0, 127.5 
Hemoglobin    
(g/dL)Mean (SD)10.1 (1.0)9.3 (1.2)<0.0001
 Median10.259.40 
 Min, Max5.6, 12.05.5, 11.8 
Ferritin    
(ng/mL)Mean (SD)31.9 (82.9)12.2 (16.2)0.0270
 Median9.856.20 
 Min, Max0.8, 881.40.9, 101.0 
Iron    
(µg/dL)Mean (SD)40.0 (28.5)36.3 (51.6)0.4761
 Median29.522.5 
 Min, Max8, 1527, 399 
TIBC    
(µg/dL)Mean (SD)339.9 (72.2)366.4 (58.1)0.0035
 Median339.5374.5 
 Min, Max175, 518174, 525 
TSAT (%)    
 Mean (SD)12.6 (9.3)10.0 (13.0)0.0732
 Median9.06.0 
 Min, Max2, 482, 96 

Correlation between screening hepcidin levels and the other screening analysts is shown in Table 3. The strongest positive correlation in nonresponders between screening hepcidin and screening iron indices was ferritin (Pearson R = 0.81, P < 0.0001). A positive correlation between screening hepcidin and screening ferritin was also observed in the responders (Pearson R = 0.59, P < 0.0001). A scatter plot for ferritin values is shown in Supporting Information Fig. S2.

Table 3. Pearson Correlation of Screening Hepcidin with Other Screening Analytes
Screening analyteDescriptive statisticNonresponder (Hgb ▵ <1 g/dL) (N = 150)Responder (Hgb ▵ ≥1 g/dL) (N = 90)Total (N = 240)
  1. Note: The percent of variability in hepcidin explained by a screening analyte can be estimated by R × R × 100 (e.g., the percent of variability in hepcidin explained by age is 0.24 × 0.24 × 100 = 5.8% for all subjects).
Age    
 Pearson R0.31−0.160.24
 P-value0.00010.13070.0002
Hemoglobin    
(g/dL)Pearson R0.010.240.12
 P-value0.94860.02410.0697
Ferritin    
(ng/mL)Pearson R0.810.590.81
 P-value<0.0001<0.0001<0.0001
Iron    
(µg/dL)Pearson R0.170.040.11
 P-value0.03300.74220.0807
TIBC    
(µg/dL)Pearson R−0.49−0.27−0.46
 P-value<0.00010.0098<0.0001
TSAT (%)    
 Pearson R0.350.080.27
 P-value<0.00010.4763<0.0001

Utilizing the hepcidin criterion of >20 ng/mL (Table 4) a sensitivity of 41.3% (62 of 150), specificity of 84.4% (76 of 90), and a PPV of 81.6% (62 of 76) could be defined. While ferritin > 30 ng/mL and/or TSAT >15% had greater sensitivity (77.3 and 64.7% respectively), their PPVs (59.2 and 55%) were inferior to hepcidin levels. Negative predictive values for hepcidin, ferritin, and TSAT were 46.3, 22.7, and 19.7 respectively.

Table 4. Predictive Values for Screening Hepcidin Levels and Iron Indices for Nonresponsiveness to Oral Iron Therapy
CriteriaPredicted responseObserved responseNumber SubjectsSensitivitySpecificityPPVNPV
Hepcidin > 20 (ng/mL)NoNo6262/150 (41.3)76/90 (84.4)62/76 (81.6)76/164 (46.3)
 NoYes14    
 YesNo88    
 YesYes76    
Ferritin > 30 (ng/mL)NoNo116116/150(77.3)10/90(11.1)116/196(59.2)10/44(22.7)
 NoYes80    
 YesNo34    
 YesYes10    
TSAT > 15 (%)NoNo9797/150(64.7)13/90(14.4)97/174(55.7)13/66(19.7)
 NoYes77    
 YesNo53    
 YesYes13    

Figure 2 illustrates the receiver operator characteristic (ROC) curve using hepcidin, ferritin, and TSAT criterion. Hepcidin was superior to both ferritin and TSAT (both of which fell below the noninformative line). Area under the curves for hepcidin, ferritin, and TSAT were 0.69, 0.40, and 0.40 respectively. Sensitivity, specificity, and positive predicting values over a range of hepcidin thresholds are shown in Supporting Information Table SII. Analyses for sensitivity, specificity, and PPV were also performed for ferritin (Supporting Information Table SIII) and TSAT levels (Supporting Information Tables SIV). Within the range of thresholds, a hepcidin criterion of >20 ng/mL had a superior PPV for predicting hemoglobin nonresponsiveness to oral iron therapy compared to ferritin or TSAT.

Figure 2.

Receiver operating characteristic curve. The ROC curves present plots of sensitivity versus (1-specificity) for hepcidin (ng/mL), ferritin (ng/mL), and TSAT (%) utilizing data from Analysis Group II at the various cutoff levels indicated near the respective curves in the same color as the respective curves. These cutoff values are also shown in Supporting Information Tables SII, SIII, and SIV, respectively.

The effect of FCM or oral iron therapy on hepcidin and hemoglobin levels from baseline (Day 1) to Day 35 was investigated in 45 of the 150 nonresponders to the oral iron run-in with these data and is summarized in Supporting Information Table SV. Patients randomized to Group A (FCM) or Group B (oral iron) had similar baseline mean hepcidin levels, (39.15 vs. 32.03 ng/mL, respectively). A greater mean change in hepcidin between Day 1 and 35 was observed in subjects randomized to FCM versus patients randomized to oral iron therapy. In the FCM group, mean hepcidin levels increased from a baseline of 39.2–160.5 ng/mL at Day 35 with a change in mean hepcidin of 121.4 ng/mL. Subjects randomized to oral iron had no significant change in hepcidin by Day 35 (32.0–29.0 ng/mL).

A multivariate logistic regression model was used to examine the simultaneous relationships between response to oral iron and screening values of hepcidin, ferritin, and TSAT. Within this multivariate model, hepcidin was statistically significant (P = 0.0025) while ferritin and TSAT were not (P = 0.7934 and P = 0.6806, respectively). This pattern indicates that ferritin and TSAT provide little incremental value beyond hepcidin.

Mean hemoglobin changes were significantly greater in subjects randomized to FCM compared to oral iron therapy (1.7 vs. 0.6 g/dL, P = 0.0025). Patients randomized to FCM had a greater percentage of subjects with a hemoglobin change ≥1 g/dL compared to oral iron (65.3 vs. 20.8%, P < 0.0001). Similar results were also demonstrated for hemoglobin change ≥2 g/dL (37.5% vs. 5.2% for FCM and oral iron, respectively).

Discussion

These analyses demonstrate significant differences in screening mean hepcidin levels between patients who are subsequently identified to be “nonresponders” to oral iron (i.e., demonstrate less than 1 g/dL increase in hemoglobin over 14 days) versus responders. We found that a hepcidin level of > 20 ng/mL showed a PPV of 81.6% for nonresponsiveness to oral iron therapy.

Hepcidin levels are influenced by a number of factors in these patients. Expression and production of hepcidin is regulated by iron status, inflammation, erythropoiesis, and oxygen tension [17]. Increased plasma iron and iron stores stimulate hepcidin production, which ultimately inhibits dietary iron absorption and decreases iron turnover. Hepcidin expression can be also stimulated by interleukin-6, an inflammatory cytokine, while increased erythropoietic activity (e.g., administration of erythropoietic stimulating agents (ESA), phlebotomy) suppresses hepcidin levels. Finally, hypoxia exerts an inhibitory effect on hepcidin production, mainly by hypoxia-inducible factor [17]. More profound iron deficiency will tend to correlate with lower hepcidin levels, but this may be modulated by induction of erythropoiesis (which also reduces hepcidin levels) as well as being modulated by pro-inflammory mediators) which elevate hepcidin levels [18].

In a previous study that investigated 44 patients with mild to moderate kidney disease being treated with ESA, hepcidin levels were found to be positively correlated with ferritin levels [19]. Similarly, hepcidin was also positively correlated with ferritin (Pearson R = 0.81, P < 0.0001) in our sample. Serum ferritin levels are affected by inflammation, in which hepcidin internalizes and degrades the ferroportin transport molecule, entrapping ferritin-bound iron with macrophages [4]. This shift to ferritin-bound iron is reflected in a corresponding increase in serum ferritin, thus termed an “acute phase reactant” to inflammation. Previous studies have shown that only when the serum ferritin cut-off is set above 30 ng/mL, will values below the cut-off predict iron deficiency with a sensitivity greater than 92% [20].

We also evaluated the change of hepcidin from baseline (Day -1) to Day 35 in patients randomized between IV iron (FCM) and oral iron therapy. We found a significant increase in hepcidin with FCM therapy but no significant change in hepcidin with oral iron therapy. An increase in hepcidin would be expected if oral iron therapy resulted in increased iron stores. That this did not occur in the oral iron group demonstrated an inferior increment in iron stores when compared to FCM therapy [21]. FCM also increased hemoglobin levels significantly when compared to oral iron therapy. The markedly superior response rates to FCM versus oral iron therapy (65.3 vs. 20.8%, respectively) confirmed the efficacy of this anemia management strategy even in patients who have previously been demonstrated to be nonresponsive to oral iron. Screening hepcidin levels can predict patients who are nonresponders, thus making oral iron therapy both undesirable and unnecessary. Finally, traditional management of IDA with the use of oral iron therapy can be misleading, since neither ferritin nor TSAT have good PPVs and nonresponsiveness to oral iron therapy does not rule out iron deficiency in patients with anemia.

Conclusion

Our analysis provides evidence that nonresponsiveness to oral iron in patients with IDA can be predicted from patients' baseline hepcidin levels, which have superior PPVs compared to TSAT or ferritin levels. Furthermore, nonresponsiveness to oral iron therapy does not rule out iron-deficiency, since IV FCM therapy produced hemoglobin responses in two-thirds of patients who had no response to a trial of oral iron therapy.

Ancillary